monsoon intraseasonal oscillations as simulated by the

Monsoon intraseasonal oscillations as simulatedby the superparameterized CommunityAtmosphere Model

Bidyut B. Goswami,1 Neena Joseph Mani,1 P. Mukhopadhyay,1 Duane E. Waliser,2

James J. Benedict,3 Eric D. Maloney,3 Marat Khairoutdinov,4 and B. N. Goswami1

Received 16 March 2011; revised 9 September 2011; accepted 9 September 2011; published 18 November 2011.

[1] The relative success of the Community Atmosphere Model with superparameterizedconvection (SP‐CAM) in simulating the space‐time characteristics of the Madden JulianOscillation encourages us to examine its simulation of the Indian summer monsoon andmonsoon intraseasonal oscillations (MISOs). While the model simulates the onset andwithdrawal of the Indian monsoon realistically, it has a significant wet bias in borealsummer precipitation over the Asian monsoon region. The space‐time characteristics ofthe MISOs simulated by the SP‐CAM are examined in detail and compared with those ofthe observed MISO to gain insight into the model’s bias in simulating the seasonal mean.During northern summer, the model simulates a 20 day mode and a 60 day mode inplace of the observed 15 and 45 day modes, respectively. The simulated 20 day modeappears to have no observed analog with a baroclinic vertical structure and strongnorthward propagation over Indian longitudes. The simulated 60 day mode seems to be alower‐frequency version of the observed 45 day mode with relatively slower northwardpropagation. The model’s underestimation of light rain events and overestimation of heavyrain events are shown to be responsible for the wet bias of the model. More frequentoccurrence of heavy rain events in the model is, in turn, related to the vertical structure ofthe higher‐frequency modes. Northward propagation of the simulated 20 day mode isassociated with a strong cyclonic vorticity at low levels north of the heating maximumassociated with a smaller meridional scale of the simulated mode. The simulated verticalstructure of heating indicates a strong maximum in the upper troposphere between 200 and300 hPa. Such a heating profile seems to generate a higher‐order baroclinic mode responsewith smaller meridional structure, stronger low‐level cyclonic vorticity, enhancedlow‐level moisture convergence, and higher precipitation. Therefore, the vertical structureof heating simulated by the cloud‐resolving model within SP‐CAM may hold the key forimproving the precipitation bias in the model.

Citation: Goswami, B. B., N. J. Mani, P. Mukhopadhyay, D. E. Waliser, J. J. Benedict, E. D. Maloney, M. Khairoutdinov, andB. N. Goswami (2011), Monsoon intraseasonal oscillations as simulated by the superparameterized Community AtmosphereModel, J. Geophys. Res., 116, D22104, doi:10.1029/2011JD015948.

1. Introduction

[2] Through determining the interannual variability of theseasonal mean, on one hand, and clustering of synopticdisturbances, on the other, northward propagating [e.g.,Yasunari, 1979; Sikka and Gadgil, 1980] monsoon intra-

seasonal oscillations (MISOs) represent major buildingblocks of the Indian summer monsoon (ISM) [Goswamiet al., 2006]. The MISO represents a signal with amplitudeas large as the annual cycle and much larger than the inter-annual variability of the seasonal mean [Waliser, 2006;Goswami et al., 2011]. The MISO manifests in the form ofactive and break spells and imparts a significant challengeand point of promise for water resource management. Hence,extended range prediction of the active and break spells hastremendous socioeconomic implications. In order for fore-cast models to have skill in predicting the active and breakspells, they must simulate the space‐time spectra of theMISO with fidelity. A space‐time spectral analysis of theIndian MISO indicates an eastward propagating componentwith periodicity between 20 and 70 days and zonal wave

1Indian Institute of Tropical Meteorology, Pune, India.2Jet Propulsion Laboratory, California Institute of Technology,

Pasadena, California, USA.3Department of Atmospheric Science, Colorado State University, Fort

Collins, Colorado, USA.4School of Marine and Atmospheric Sciences, Stony Brook University,

New York, New York, USA.

Copyright 2011 by the American Geophysical Union.0148‐0227/11/2011JD015948

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, D22104, doi:10.1029/2011JD015948, 2011

D22104 1 of 17

http://dx.doi.org/10.1029/2011JD015948

numbers between 1 and 3, which is about 1.5 times strongerthan the corresponding westward propagating component[see Lin et al., 2008, Figure 10]. The MISO over the Indianregion is also characterized by strong northward propagationthat is about 4 times stronger than the correspondingsouthward propagating component [see Lin et al., 2008,Figure 15a]. However, most climate models exhibit diffi-culty in simulating the MISO [Waliser et al., 2003; Lin et al.,2008] and underestimate the northward propagating com-ponent, in particular [see Lin et al., 2008, Figure 15a].Because of their contribution to the seasonal mean, improvedMISO simulation in a model would have a beneficial impactnot only on the extended range prediction of the active andbreak spells, but also in seasonal prediction.[3] As the problem of better simulation of the MISOs in

climate models is similar to the difficulties faced in simu-lating the Madden Julian Oscillation (MJO), we may gaininsight from the experience of the community in attemptingto improve MJO simulations. Considerable effort has beenexpended during the last few years to improve the poorsimulation of the MJO by models [e.g., Slingo et al., 2005;Lin et al., 2006; Waliser et al., 2009; Kim et al., 2009]. Inaddition to efforts to improve the MJO through modifica-tions to conventional convection parameterizations, methodsinvolving explicit representation of some aspects of cloudshave emerged to improve the poor simulation of the MJO.For example, global cloud‐resolving model simulationshave been conducted that produce realistic mesoscale cloudclusters and have demonstrated better simulation of the MJO[Miura et al., 2007; Sperber et al., 2008; Oouchi et al., 2009;Liu et al., 2009; Sato et al., 2009]. However, due to the largecomputation cost of such models, the ability to simulatemultiple years to analyze systematic biases in the MJO lifecycle remains computationally prohibitive [Khairoutdinovet al., 2008, hereafter KDR08]. Another rather attractivealternative to global‐cloud resolving models is the multi-scale modeling framework (MMF) [Khairoutdinov andRandall, 2001; Khairoutdinov et al., 2005; Tao et al.,2009] that has demonstrated good simulation of MJOspace‐time characteristics [Benedict and Randall, 2009].[4] In the MMF approach, a 2‐D cloud‐resolving model

(CRM) is embedded with each grid column of a general cir-culation model (GCM) such that explicit simulation of sub-grid scale cloud processes replaces the traditional cumulusparameterization of the GCM, hence minimizing the need forimperfect parameterization assumptions. Because the CRMtime step is much shorter than that of the GCM, the CRM isintegrated several times in oneGCM time step (for details, seeBenedict and Randall [2009]). The horizontal average of theCRM output within a GCM grid box determines the physicaltendencies for the next GCM time step.[5] While the MMF approach has been extensively tested

for simulating the eastward propagating convectionanomalies from the Indian ocean to the western Pacificduring boreal winter [Benedict and Randall, 2009], rela-tively little work has been conducted to address the fidelityof the MMF in simulating the northward propagating ISMintraseasonal oscillation [Stan et al., 2010; DeMott et al.,2011]. As the MISO and ISM forecasting remain a chal-lenge for both statistical and dynamical approaches, and as ithas significant societal impacts, the objective of this paper isto investigate the quality of the MISO simulation as obtained

from an atmospheric model intercomparison project (AMIP)run of the SP‐CAM. Such an analysis and comparison withthe observedMISOmay suggest the utility of the SP‐CAM tobe used as a tool for extended range prediction of the MISOor as a “laboratory” to better understand shortcomings inconvection‐parameterized models. In addition, as the MISOcontributes significantly to the seasonal mean precipitationand winds, a diagnosis of the biases in the MISO simulationmay also provide clues toward improvement of the simula-tion of the seasonal mean climate.[6] The model and data used to diagnose the model are

described in section 2, and simulation of the monsoon cli-matology is described in section 3. Basic characteristics ofthe monsoon intraseasonal variability (ISV) as produced bythe model are presented in section 4, and a more detaileddiagnosis of the simulated variability is made in section 5.This is followed in section 6 by a discussion of how biasesin simulating the MISO affect the seasonal mean bias. Majorconclusions are summarized in section 7.

2. Model Simulations and Data Used

[7] The National Center for Atmospheric Research(NCAR) Community Atmosphere Model version 3 (CAM3.0) [Collins et al., 2006] acts as the host GCM for the super-parameterization approach employed [e.g., Khairoutdinovet al., 2005]. The version of CAM3.0 used has a 2.8° × 2.8°horizontal grid (T42 spatial truncation), 30 vertical levels upto 3.6 hPa, and a time step of 30 min. Embedded within eachGCM grid cell is a 2‐D CRM composed of 32 columnsoriented in the north‐south direction having 4 km horizontalgrid spacing, periodic boundary conditions, 28 vertical levelscollocated with the 28 lowest CAM levels, and a time step of20 s. This embedded CRM effectively replaces the CAM’sconventional parameterizations of moist physics, convection,turbulence, and boundary layer processes. However, cou-pling between the surface and atmosphere is computed onlyon the GCM grid. The effects of subgrid scale temperatureand wind perturbations (e.g., gust fronts) on surface fluxesare thus not included. CRM scale enhancements of surfacedrag related to localized gustiness of near‐surface winds areexplicitly included, however. The AMIP simulation wasconducted using prescribed monthly mean (interpolated todaily mean) sea surface temperatures and sea ice concentra-tions [Hurrell et al., 2008]. The simulation produced 19 yearsof global daily output spanning 1 September 1985 to 25September 2004.[8] Further implementation details on the super-

parameterization approach can be found in the studies byBenedict and Randall [2009], Khairoutdinov and Randall[2003], and Khairoutdinov et al. [2005]. Additional detailsof the SP‐CAM AMIP simulation used in this study can befound in KDR08.[9] To evaluate the simulated precipitation, the Global

Precipitation Climatology Project (GPCP) rainfall data[Huffman et al., 2001] regridded to the model resolution(2.8° × 2.8°) are used. The regridding is done using abilinear interpolation technique. The GPCP data used formodel validation are from 1997 to 2008. The NOAA Out-going Longwave Radiation (OLR) data set is used to provideadditional diagnostics of convective variability. Horizontalwinds are taken from National Centers for Environmental

GOSWAMI ET AL.: MISO AND SP‐CAM D22104D22104

2 of 17

Prediction (NCEP)/NCAR reanalysis at 2.5° × 2.5° resolu-tion [Kalnay et al., 1996]. The NCEP winds and NOAAOLR data are used for 1986 to 2004.[10] In order to extract MISOs from the simulation and

observations, daily anomalies are calculated as the departureof daily values from a smoothed climatology at daily reso-lution. The smoothed climatology is reconstructed based onthe annual mean and first three harmonics of the long‐termmean seasonal cycle.

3. Simulation of Monsoon Climatology

[11] AsMISOs are hypothesized to arise from a convective‐radiative‐dynamical feedback process [e.g., Goswami andShukla, 1984; Wang, 2005], they critically depend on the

background climatic state. Therefore, examination of thesimulated climatological mean field is necessary for under-standing any biases in the model MISO simulation. In thissection, we present an overview of simulated climatologicalmean June–July–August–September (JJAS) precipitationand winds and compare them with observations.[12] As reliable high‐resolution rainfall data are available

over continental India compiled by the India MeteorologicalDepartment (IMD), [Rajeevan et al., 2006], we compare theclimatological mean JJAS precipitation simulated by SP‐CAM with IMD data (Figures 1a and 1b). It is noted that themodel simulates the climatological rainfall distribution overthe continent reasonably well with a pattern correlation of0.56, except that the rain shadow region over southeast Indiais rather small in the model simulation. This appears to be

Figure 1. Climatological mean (JJAS) precipitation (mm d−1) from (a) IMD and (b) SP‐CAM. Standarddeviation of 10–90 day bandpass‐filtered daily precipitation anomalies (mm d−1) (JJAS) from (c) IMDand (d) SP‐CAM.


3 of 17

related to the inability of the climate model to resolve andsimulate the amplitude of the Western Ghat precipitation.The amplitude of ISV as represented by the standard devi-ation of 10–90 day filtered rainfall anomalies simulated bythemodelwithin the continent is reasonable (Figures 1c and 1d)with a pattern correlation of 0.56, except that the model ISVis stronger than observed over east‐central India and weakerthan observed close to the Western Ghat.[13] The Indian monsoon, however, is not confined within

continental India and has a much larger spatial scale[Webster et al., 1998]. To document the model biases insimulating the Indian monsoon on a larger scale, a compar-ison of the climatological mean JJAS precipitation in the SP‐CAM and the observations using GPCP is made (Figure 2).While it indicates some similarities, a few important differ-ences are noted. First, the northern rainband or the conti-nental tropical convergence zone (TCZ) extends northwardto about 20°N in both the SP‐CAM simulations (Figure 2b)and observations (Figure 2a). Also, as in observations, themodel simulates a maximum in precipitation in the equatorialIndian Ocean with a dry zone between the oceanic maximumor the oceanic TCZ and the continental TCZ. Furthermore,the oceanic TCZ is located in the eastern Indian Ocean as inobservations. However, the SP‐CAM simulation is far toowet in the regions of maximum precipitation compared toobservations (Figure 2b). Also, it simulates an extra center of

high precipitation in the western Bay of Bengal (BoB), nearthe eastern coast of India at 15°N. Another feature of thesimulation is that the northern rainband is too zonally ori-ented compared to the northeast‐to‐southwest orientation ofthe observed precipitation (Figure 2a).[14] The strong wet bias and zonally oriented nature of

simulated precipitation are also reflected in biases of thesimulated JJAS climatological mean winds at 850 hPa(Figure 2d) compared to the observed climatology (Figure 2c).The low‐level jet (LLJ) at 850 hPa is much stronger in themodel than in observations, consistent with the model wetbias. Furthermore, consistent with the zonally oriented char-acter of the simulated precipitation (Figure 2b), the windsassociated with the LLJ are too zonally oriented, and strongzonal winds extend too far into the western Pacific. As aresult, the climatological vorticity at 850 hPa simulated by themodel is also too zonal and fails to simulate the cyclonicvorticity maximum in southeastern India (Figure 2c).[15] In order to examine whether the model correctly si-

mulates the onset and withdrawal of the monsoon and thetime evolution of rainfall over the continent and near‐coastalocean, the annual cycle of daily climatological rainfallaveraged over the two boxes in Figure 2b is computed(Figures 3a and 3b). The fluctuations in the daily climatol-ogy of rainfall from GPCP as well as from SP‐CAM seen inFigures 3a and 3b are the climatological intraseasonal os-

Figure 2. Climatological mean (JJAS) precipitation (mm d−1) from (a) GPCP and (b) SP‐CAM and850 hPa winds (m s−1) from (c) NCEP and (d) SP‐CAM. In Figures 2c and 2d, the shaded area is thecorresponding vorticity (10−6 s−1), and the thick white line is the zero vorticity.


4 of 17

cillations (CISO) [Wang and Xu, 1997]. However, becauseof the limited number of years (12) to create the clima-tology for GPCP, some sampling bias may be present inthe GPCP CISO. The smoothed climatology constructedfrom the first three harmonics of the daily climatology andannual mean for both GPCP and SP‐CAM are also shownin Figure 3. It is encouraging to note that the model sim-ulates a rapid onset and a slow withdrawal of monsoonalprecipitation as observed [Krishna Kumar et al., 2011].The timing of these onsets and withdrawals are similar toobservations, with a rapid increase in climatological pre-cipitation during May and gradual decline during Augustand September. A simulated wet bias persists throughoutthe summer monsoon season, however. It is also encour-aging to note that during the winter season, the simulatedrainfall is near zero as in observations. The annual cycleand northward migration of the TCZ is explored byplotting simulated daily climatological precipitation aver-aged from 70°E and 90°E as a function of latitude andtime (Figure 4b) compared to observations (Figure 4a). It isnoted that the model shows promise in simulating thedouble TCZ in the monsoon domain (Figure 4b), one overthe Indian subcontinent and the other over the IndianOcean, owing to a reasonable northward propagation of theTCZ. The model, however, shows a wet bias over the

Figure 3. Seasonal evolution of rainfall (mm d−1) overthe two boxes, (a) Box‐1 and (b) Box‐2, shown inFigure 2b. Smoothed climatology is on top of unsmootheddaily climatology.

Figure 4. Climatological annual cycle of precipitation over70°–90°E (mm d−1) from (a) GPCP and (b) SP‐CAM.

Figure 5. Standard deviation of 10–90 day bandpass‐filtered daily precipitation (mm d−1) anomalies (JJAS) from(a) GPCP and (b) SP‐CAM. (The boxes in Figure 5b areplotted for Figures 15 and 16.)


5 of 17

continental branch of the TCZ, which increases in amplitudeearlier than observations (Figure 4a).

4. Simulation of Monsoon IntraseasonalVariability

[16] The space‐time characteristics of simulated monsoonintraseasonal variability are examined in this section andcompared with those of the observed variability. We definethe amplitude of ISV as the standard deviation of the 10–90 day bandpass‐filtered daily precipitation anomaly duringJune–September. Model variability (Figure 5b) is comparedto that from observations (Figure 5a). The spatial patterns ofprecipitation ISV in observation and the model (Figure 5)are similar to the climatological seasonal mean precipitation

(Figure 2a). This relationship between the mean and vari-ance appears to be consistent with the fact that observed andsimulated precipitation follows a Poisson distribution (figurenot shown) for which variance is proportional to the mean.Thus, compared to observations, the stronger ISV simulatedby the model is consistent with its large wet bias in thesimulation of seasonal mean precipitation.[17] A quick assessment of the model’s ability to simulate

the eastward and northward propagation characteristics ofsummer ISV over the Indian continent is done in Figure 6.This is based on lag regressions of 10–90 day filtered pre-cipitation with respect to a reference time series of 10–90 dayfiltered precipitation averaged over central India. It is seenthat within the continental India, the model simulates theeastward (Figure 6b) and northward (Figure 6d) propagation

Figure 6. (a) Longitude‐time and (c) latitude‐time plots of 10–90 day filtered IMD anomalies regressedon CIPI, averaged over 10°–20°N and 70°–90°E, respectively. (b and d) Same as Figures 6a and 6c butfor SP‐CAM output.


6 of 17

reasonably well compared to IMD observations (Figures 6aand 6c). However, it is consistent with earlier discussionsof the model’s inability to simulate the observed high ISV inthe Western Ghat region (Figures 6a and 6b).[18] In order to examine the model’s fidelity in simulating

the dominant modes of ISV during northern summer, space‐time spectra of daily precipitation from GPCP and from theSP‐CAM are calculated following the methodology ofWheeler and Kiladis [1999]. The daily anomalies from theJJAS period for all 12 available years of GPCP and a similar12 years of data from the SP‐CAM are used in the spectralcalculations. Space‐time spectra are calculated for zonallypropagating modes using global data at each latitude andthen averaging the resulting power between 20°S and 30°N.These spectra are shown for GPCP (Figure 7a) and themodel (Figure 7b). Similarly, space‐time spectra formeridional propagation are calculated using data between20°S and 30°N and then averaging the spectral powerbetween 60°E and 135°E. These spectra are shown inFigure 7c from GPCP and in Figure 7d from the model. Incalculating the space‐time spectra for meridional propaga-tion, the choice of the domain bounds is based on the factthat the northward propagating MISO is generally confinedbetween 20°S and 30°N within this domain [see Goswami,2005]. Meridional wave number 1 is defined on the basisof the largest wave that entirely fits inside the domain. In

interpreting these spectra, it must be kept in mind that thedomain is not periodic and hence some artificial spectralpower may be introduced into the wave number‐frequencyspectrum. However, sensitivity tests indicate that the loca-tion and power of the dominant spectral signals are insen-sitive to modest changes in the domain size.[19] The spectra indicate the following behavior.[20] 1. The model simulates the eastward propagating

mode at wave number 1 with a longer period (60 days) thanin observations (45 days). With significant power at west-ward propagating wave number 1, the simulated mode mayhave a stationary component, although we would need toconduct an analysis of coherence between eastward andwestward components to confirm this.[21] 2. In observations, the dominant northward propa-

gating mode appears to be the same 45 day eastwardpropagating mode (Figures 7a and 7c). However, in themodel, the dominant northward propagating mode is a 20 daymode with no zonal propagating counterpart (Figures 7band 7d). The simulated 60 day mode, in addition to havinga meridional structure with meridional wave number 1, has asignificant power in meridional mean component.[22] Thus, in addition to simulating the observed 45 day

mode with a longer period of 60 days, the model’s dominantnorthward propagating mode in the region is a 20 day mode.A question that can be raised is the degree to which this new

Figure 7. Wave number‐frequency power spectra of precipitation. Zonal power spectra (divided by thebackground) calculated over 0°–360°, 15°S–15°N for (a) GPCP precipitation and (b) model‐simulatedprecipitation. Meridional power spectra calculated over 15°S–30°N, 60°–110°E for (c) GPCP precipita-tion and (d) model‐simulated precipitation.


7 of 17

Figure 8. Wave number‐frequency power spectra of zonal wind at 850 Mb. Zonal power spectra calcu-lated over 0°–360°, 15°S–15°N for (a) NCEP/NCAR reanalysis U wind and (b) model‐simulated U wind.Meridional power spectra calculated over 15°S–30°N, 60°–110°E for (c) NCEP/NCAR reanalysis U windand (d) model‐simulated U wind.

Figure 9. Space‐time coherence squared spectrum for symmetric latitudes 15°S–15°N: (a) model‐simulated precipitation with zonal wind at 850 hPa; (b) GPCP precipitation with NCEP/NCARreanalysis wind at 850 hPa; (c) model‐simulated zonal wind at 200 hPa with zonal wind at 850 hPa;and (d) NCEP/NCAR reanalysis zonal wind at 200 hPa with that at 850 hPa.


8 of 17

model mode exhibits coherence between convection and thelarge‐scale circulation. Comparing the space time spectra ofdaily 850 hPa (JJAS) winds simulated by SP‐CAM with thespace‐time spectra of observed winds (Figure 8), we findthat the meridional propagating component of the simulated60 day mode exhibits coincident peaks in wind and pre-cipitation, indicating strong coupling between convectionand the large‐scale circulation. While the simulated 20 daymode in precipitation (Figure 7d) clearly exhibits northwardpropagation, near the same periods in low‐level winds, asignificant stationary component is indicated (Figure 8d)with comparable power in both northward and southwardwave number bands.[23] To gain further insight into the covariability of the

model convection and circulation fields in the spectral

domain, the coherence squared spectra between symmetricprecipitation and winds were estimated using the dailyanomalies. The coherence squared spectrum is obtained bynormalizing the cross‐power spectrum that contains bothmagnitude and phase of the relationship between the twofields, with the individual powers in each field [Hendon andWheeler, 2008]. It can be interpreted as the wave number‐frequency distribution of the squared correlation coefficientbetween the two fields. The coherence squared spectrabetween the SP‐CAM symmetric precipitation and dailywind fields at 850 hPa (Figure 9a) shows a peak coherencesquared of 0.4 coherence associated with the MJO that isweaker than the coherence exhibited by GPCP precipitationand NCEP winds at 850 hPa (Figure 9b). The model sym-metric precipitation and 200 hPa wind fields show very high

Figure 10. (a) Longitude‐time and (c) latitude‐time plots of 10–20 day filtered GPCP anomaliesregressed on CIPI, averaged over 10°–20°N and 70°–90°E, respectively. (b and d) Same as Figures 10aand 10c but for SP‐CAM output with 15–30 day filter.


9 of 17

coherence corresponding to the MJO (not shown). While theobserved precipitation field (GPCP) exhibits strong coher-ence with both lower and upper level circulation, the modelprecipitation field is more strongly coupled with the upperlevel winds, indicating a bias in the vertical wind structure.The small coherence squared spectra of 200 and 850 hPasymmetric model winds in the MJO band (Figure 9c) alsocorroborates this notion, in contrast to the strong correlationbetween upper and lower level winds seen in observations(Figure 9d).

5. Diagnosis of the Simulated Modes

[24] Observed monsoon intraseasonal variability consistsof a well‐documented quasi‐biweekly mode (or a 15 daymode) as described in many studies and a mode centered on45 days (for a review, see Goswami [2005]). These modeshave distinct horizontal and vertical structures and propa-gation characteristics. The simulated 20 and 60 day modesdo not occupy the same place in the space‐time spectra asthe observed modes, and so it is important to identify towhat extent the model modes are dynamically distinct fromthe observed modes and what the simulated modes repre-sent. For example, is the simulated 20 day mode similar tothe observed 15 day mode, and is the simulated 60 daymode simply a model version of the observed 45 daymode?

5.1. The 20 Day Mode

[25] We first compare horizontal and vertical structure andpropagation characteristics of the simulated 20 day modeand the observed 15 day mode. For this purpose, the sim-ulated 20 day mode in precipitation and winds is isolatedusing a 15–30 day Lanczos filter [Duchon, 1979]. Theobserved 15 day mode is isolated by using a 10–20 dayLanczos filter on GPCP daily precipitation and NCEP windanomalies. We use lag‐regression techniques to illustratepropagation characteristics of these modes for which a ref-erence time series is needed [Wilks, 1995]. A referenceprecipitation time series for use in regression is created byaveraging filtered anomalies over central India between15°–25°N, 73°–82°E where the amplitude of variability islarge both in observation and in model simulation. We thenlag regress filtered anomalies for the entire map on to thereference time series. The zonal propagation of the respec-tive modes is illustrated in Figures 8a and 8b, whereregression coefficients averaged between 10°N and 20°Nare plotted. Similarly, north‐south propagation is illustratedin Figures 10c and 10d where the regressed values averagedbetween 70°E and 90°E are plotted. While the observed15 day mode has clear westward propagation (Figure 10a), thesimulated 20 day mode is relatively stationary in the east‐west direction (Figure 10b). Although there seems to betwo discontinuous precipitation regimes, one east of 100°Eand another west of this line, in both the regimes, the sim-ulated mode does not have a strong propagating character.

Figure 11. Filtered 200 hPa wind anomalies regressed on CIPI: (a) 10–20 day filtered NCEP; (b) 30–80 day filtered NCEP; (c) 15–30 day filtered SP‐CAM; and (d) 35–80 day filtered SP‐CAM. Vorticityanomalies are in colors.


10 of 17

Similarly, while the observed 15 day mode is relativelystationary in the north‐south direction (Figure 10c), thesimulated 20 day mode is clearly northward propagating at arate of about 1.2° d−1 (Figure 10d).[26] In order to gain insight into the vertical structure of

the simulated 20 day mode and compare it with the verticalstructure of the observed 15 day mode, filtered windanomalies are regressed onto the filtered central Indiaprecipitation index (CIPI). The regressed wind patterns forthe observed 15 day mode (Figures 11a and 12a) show thatthe 850 hPa cyclonic anomaly pattern (Figure 12a) alsoextends up to 200 hPa (Figure 11a). This barotropic verticalstructure of the observed 15 day mode is consistent withearlier studies [Chatterjee and Goswami, 2004]. For thesimulated 20 day mode, however, the cyclonic and con-vergent 850 hPa circulation (Figure 12c) is overlaid by adivergent anticyclonic circulation pattern at 200 hPa(Figure 11c), indicating that together with a barotropiccomponent, a baroclinic component may be present in thevertical structure. A more detailed examination of thevertical structure associated with the mode (see Figure 14b)also supports this conclusion.[27] Although the period of the simulated 20 day mode is

close to that of the observed 15 day mode, the verticalstructure and propagation characteristics of the mode sug-gest that it is distinctly different from the observed 15 daymode. Thus, it may be inferred that the simulated 20 daymodel is a new model‐generated mode that is not observed.

5.2. The 60 Day Mode

[28] The simulated 60 day mode, as well as the observed45 day mode in precipitation and winds, is isolated using a30–80 day (for observation) and a 35–80 day (for simula-tion) Lanczos filter. A reference time series is created byaveraging filtered anomalies over central India between15°–25°N, 73°–82°E for GPCP and for the model‐simulatedrainfall. Lag regressions of the filtered anomalies are thencalculated with respect to the reference time series. Thezonal propagation of the mode is illustrated in Figures 13aand 13b where regressed values averaged between 10°Nand 20°N are plotted. Similarly, meridional propagation isillustrated in Figures 13c and 13d where the regressedvalues averaged between 70°E and 90°E are plotted. Further-more, while the observed 45 day mode has a clear eastwardpropagation (Figure 13a), the simulated 60 daymode has onlya weak eastward propagation between 60°E and 120°E(Figure 13b). Similarly, while the observed 45 daymode has aclear northward propagation with speed of about 1.1° lat d−1

(Figure 13c), the simulated 60 day mode exhibits muchslower northward propagation at a rate of about 0.5° lat d−1

(Figure 13d).[29] Filtered wind anomalies are regressed against the

filtered precipitation index averaged over central India (10°–20°N, 70°–90°E) to examine the vertical structure of thesimulated and observed modes. The regressed wind patternsfor the observed 45 day mode (Figure 12b) show that the850 hPa cyclonic and convergent pattern of anomaly is

Figure 12. Filtered 850 hPa wind anomalies regressed on CIPI: (a) 10–20 day filtered NCEP; (b) 30–80 day filtered NCEP; (c) 15–30 day filtered SP‐CAM; and (d) 35–80 day filtered SP‐CAM. The cor-responding vorticity (10−6 s−1) is plotted in the background (shading). Line plots (middle top andbottom) are correspondingly filtered regressed vorticity (10−6 s−1) averaged over 70°–85°E; the blueline represents NCEP output, and the red line represents SP‐CAM output.


11 of 17

overlaid by a divergent anticyclonic anomaly pattern at200 hPa (Figure 11b). The simulated 60 day mode alsodemonstrates a similar cyclonic and convergent 850 hPacirculation overlaid by a divergent anticyclonic circulationpattern at 200 hPa (Figures 12d and 11d).[30] Both the observed 45 day mode and the simulated

60 day mode have a similar dominant baroclinic compo-nent in the vertical structure and time scale. Thus, thesimulated 60 day mode seems to be the model version ofthe observed 45 day mode. Comparing Figures 12b and12d, we find that the horizontal scale of simulated 60 daymode (red line) is smaller than that of the observed 45 daymode (blue line). It can also be noted that the simulatedregressed wind field at the 20 day mode (Figure 12c)

resembles that of the simulated 60 day mode (Figure 12d),particularly over the monsoon region.

5.3. On the Northward Propagation

[31] It has been suggested that the northward propagationof the observed 45 day mode is a result of a feedbackbetween convective heating and circulation [e.g., Jianget al., 2004]. In the presence of an easterly shear of back-ground winds, the atmospheric response to a heatinganomaly produces a barotropic cyclonic vorticity center andanomalous lower tropospheric moisture convergence to thenorth of the heating maximum. This helps the heatinganomaly to propagate northward. To examine the model’sfidelity in simulating this process, bandpass‐filtered vortic-ity anomalies are composited relative to OLR anomalies

Figure 13. (a) Longitude‐time and (c) latitude‐time plots of 30–80 day filtered GPCP anomaliesregressed on CIPI, averaged over 10°–20°N and 70°–90°E, respectively. (b and d) Same as Figures 13aand 13c but for SP‐CAM output with 35–80 day filter.


12 of 17

corresponding to active spells, defined using minima offiltered OLR averaged over central India (15°–25°N, 73°–82°E). The vertical structure of the vorticity anomalies forthe simulated 20 and 60 day modes averaged between 70°Eand 90°E are shown in Figures 14b and 14d as a function oflatitude. Similar composites of vorticity anomaly for theobserved 15 day mode and the 45 day mode are shown inFigures 14a and 14c. The composite OLR for each modeaveraged between 70°E and 90°E are also shown beloweach vorticity plot, in addition to the anomalous 850 hPavorticity. It is noted that the minimum of OLR (convectionmaximum) for the observed 15 day mode is nearly collo-cated with the location of the barotropic vorticity maximum(Figure 14a, top and bottom), consistent with the lack ofnorthward propagation in this case as in the theory of Jianget al. [2004]. However, the low‐level cyclonic vorticitymaximum for the model 20 day mode is slightly shiftednorth of the OLR minimum (Figure 14b, top and bottom),consistent with the northward propagation of this mode.Consistent with Figures 12c and 11c, in the case of thesimulated 20 day mode, the center of low‐level cyclonicvorticity is overlaid by an anomalous anticyclonic circula-tion in the upper troposphere, indicating the baroclinicnature of the vertical structure.[32] In the case of the simulated 60 day mode (Figure 14d,

top and bottom), however, the low‐level vorticity maximumis nearly collocated with the OLR minimum, consistent witha weak northward propagation or a nearly stationary char-acter of this mode in the north‐south direction (Figure 13d).For the observed 45 day mode (Figure 14c, top and bottom),

the cyclonic vorticity maximum is located north of the OLRminimum, as expected.

6. Diagnostic of Seasonal Mean Bias

6.1. Simulated Probability Distribution of DailyRainfall and the Seasonal Mean Bias

[33] The analyses in the earlier sections have brought outthe wet bias of the model compared to observation based ondaily rainfall climatology. As the mean and variance ofrainfall distribution are linearly related, it is possible that thebias in simulating the mean rainfall may be related to biasesin simulating the frequency distribution of rain rates. Threerepresentative areas over central India (Figure 5b, box 1),Bay of Bengal (Figure 5b, box 2), and Arabian Sea (Figure5b, box 3) are chosen, and the probability distributionfunction (PDF) in percentage is computed for different rainrate categories, based on daily rainfall with a bin width of5 mm. The PDF over central India (Figure 15a) reveals thatthe model underestimates the probability of occurrence oflighter rain rates and overestimates the moderate rain rates.The observations over Bay of Bengal (Figure 15b) andArabian Sea (Figure 15c) show that lighter rain rate cate-gories dominate, which the model is not able to capture.The model rain rates are shifted toward moderate andheavy categories over both the oceanic basins. Thus, themodel shows a tendency for simulating heavy rain eventswith higher frequency over oceanic regions and also, to acertain extent, over Central India and fails to simulate thelighter rainfall categories with the same frequency as in the

Figure 14. Composite of vorticity (5 × 10−6 s−1) anomaly averaged over 70°–90°E, for the activemonsoon condition, for observed data (NCEP) with (a) 10–20 day filter and (c) 30–80 day filterand for SP‐CAM output with (b) 15–30 day filter and (d) 35–80 day filter. Corresponding OLR(Wm−2) (red line) and 850 hPa vorticity (blue line) composites are plotted at the bottom.


13 of 17

observations. These biases explain the wet bias in thesimulation of the seasonal mean rainfall.

6.2. Vertical Structure of Rain Events and SeasonalMean Bias

[34] The above analyses clearly demonstrate the modeltendency of generating more heavy rainfall events, whichseems to suggest too frequent deep convection. Thus, anatural question is what could be the possible source of suchmodel bias? To find an answer to this question, we want tofurther quantify the possible source of such model behaviorby analyzing the vertical structure of heating.[35] A significant fraction of monsoon interannual vari-

ability (IAV) in observations as well as in models arise from“internal dynamics” and appears to owe its origin to theintraseasonal oscillations [Goswami, 1998; Ajaya Mohanand Goswami, 2003; Goswami and Xavier, 2005]. Mon-soon intraseasonal oscillations (ISOs) may influence theseasonal mean monsoon through two different mechanisms.As the spatial structure of the dominant MISO has a sig-nificant projection on the dominant mode of IAV of themonsoon, an asymmetry in the frequency of occurrence ofactive and break spells can lead to a bias of the seasonalmean ISO anomalies leading to a contribution to the sea-sonal mean [Goswami et al., 2006]. However, this linearmechanism cannot explain the total internal IAV simulatedby a model [Goswami and Xavier, 2005]. Also, the non-linear relationship between area averaged variance of ISO

anomalies and IAV of seasonal mean monsoon [Goswamiand Xavier, 2005] indicate that there exists a nonlinearmechanism through which the MISO could influence theseasonal mean. Therefore, it is reasonable to think thatbiases of a model in simulating the seasonal mean monsoonmay be linked to the biases of the model in representing theMISO.[36] From Figure 14, we note that the simulated cyclonic

vorticity anomalies associated with active phases of both thelower‐ and higher‐frequency modes are more than threetimes stronger than the observed anomalies. In particular,the cyclonic vorticity anomalies associated with the simu-lated 20 day mode are even stronger than those for thesimulated 60 day mode. Therefore, the simulated 20 daymode not only dominates the ISV, it may also influence theseasonal mean. Another interesting feature about the verticalstructure of the vorticity anomaly associated with the twosimulated modes is that unlike the observed modes wherethe maximum of vorticity anomaly tends to occur aloft, themaximum vorticity anomalies for the simulated modes occurvery near the surface (Figure 14).[37] The low‐level cyclonic vorticity simulated by the

model produces strong low‐level moisture convergence andleads to the positive precipitation bias of the model (seeFigure 2). As mentioned earlier, since the ISOs can influ-ence the seasonal mean, a strong 20 day mode in the modelwith its associated strong vorticity may be a leading con-tributor to strong low‐level moisture convergence and themean precipitation bias of the model. Therefore, it isimportant to understand why the vorticity anomalies in thesimulated modes are generally much stronger than thoseobserved and why it is particularly strong in the 20 daymode. One possibility is that the meridional scale of thesimulated modes is significantly smaller than that of theobserved modes (see Figure 13), leading to a much strongermeridional gradient of zonal wind anomalies and strongervorticity anomaly. Figure 14b suggests that the verticalstructure of vorticity anomalies associated with the 20 daysimulated mode is characterized by a higher‐order barocliniccomponent that makes it more complex than the observedmode. Dynamically, if the simulated mode is associatedwith a higher‐order baroclinic structure compared to ob-servations, it would also be associated with a smaller hori-zontal scale [Pedlosky and Frenzen, 1980]. This can happenif the heating profile produced by the model has a strongmaximum in the upper troposphere (200–300 hPa), and ifthat maximum is much stronger in the case of the 20 daymode compared to the 60 day mode. To gain some insighton this issue, we examined the vertical profiles of CRMheating tendencies for the two modes over three represen-tative regions shown in Figure 5b. The composite verticalprofile of CRM heating tendency for 15–30 and 35–80 dayfiltered data during active phases over different subregionsis shown in Figure 16. The vertical profile of heating ten-dency for active periods is found to have sharper verticalstructure for the 15–30 day mode over the BoB (Figure 16b,red line), Western Ghat (Figure 16c, red line), and centralIndian (Figure 16a, red line) regions compared to the 35–80 day mode (Figure 16, blue line). Similar differences inthe vertical profile of CRM heating tendency are also notedbetween 15 and 30 day and 35–80 day modes during breaks(not shown). It is also noted that the maximum heating rates

Figure 15. Probability distribution function in percentagefor (a) central India, (b) Bay of Bengal, and (c) ArabianSea (the boxes are shown in Figure 5b), based on dailyrainfall (mm d−1)with a bin width of 5 mm.


14 of 17

are in the upper troposphere between 200 and 300 hPa. TheCRM heating tendencies are stronger in the 15–30 daymode, and, as a response, the SP‐CAM shows enhancedupper (lower) level divergence (convergence), which adds tothe larger precipitation bias in the 15–30 day mode com-pared to the 35–80 day mode, although depending on theenergetics of the system, the large‐scale circulation biasesmay simply be a response to the precipitation bias ratherthan a cause.

7. Conclusions

[38] Realistic simulation of the observed space‐timecharacteristics of MISOs is important for a model’s ability tosimulate the seasonal mean with fidelity. In the presentstudy, we examine simulation of seasonal mean borealsummer monsoon characteristics and MISOs from a 20 yearSP‐CAM AMIP style forced by observed sea surface tem-peratures. The SP‐CAM has a reasonable seasonal cycle ofthe Indian monsoon. However, the model shows a strongwet bias during northern summer over the Asian monsoonregion. A similar wet bias in Asian summer monsoon sim-ulation by SP‐CAM was found by Luo and Stephens [2006],which they had attributed to the periodic boundary conditionof the CRMs. An attempt is made to diagnose the cause ofthis wet bias in the model simulations.[39] The model shows limited ability to capture observed

intraseasonal modes with fidelity. The model simulates aneastward propagating mode of wave number 1 with a longerperiod (60 days) compared to observations (45 days). Thesimulated mode also appears to have significant power atwestward propagating wave number 1, suggesting a sta-tionary component compared to observations. While thevertical structure of the simulated mode is similar to theobserved 45 day mode, its northward propagation is muchslower than its observed counterpart.

[40] The model does not realistically simulate theobserved quasi‐biweekly mode. Instead, it produces a verystrong 20 day mode with robust northward propagation. Thesimulated 20 day mode has a baroclinic component in thevertical structure in contrast to the observed 10–20 daymode, which has a strong barotropic component. Thus, the20 day simulated mode shows no observational analog andmay hold a key for understanding its seasonal mean pre-cipitation bias in the model.[41] The strong northward propagation of the simulated

20 day mode is associated with strong cyclonic vorticityproduced at low levels to the north of the heating maximum,consistent with the hypothesis of Jiang et al. [2004]. Strongvorticity associated with the mode, in turn, is associatedwith a smaller meridional scale of the simulated mode. Asthe circulation response depends crucially on the verticalstructure of convective heating, we examined the CRMheating rates within the SP‐CAM. Analysis of the verticalstructure of the SP‐CAM heating rates indicates that itproduces a strong heating maximum in the upper atmo-sphere between 200 and 300 hPa. Such a heating profilewould tend to support a higher‐order baroclinic moderesponse (second baroclinic mode or higher). Higher‐orderbaroclinic modes are generally associated with a smallermeridional structure, stronger low‐level cyclonic vorticity,stronger low‐level moisture convergence, and higher pre-cipitation. The vertical structure of heating simulated by theCRM within SP‐CAM may therefore hold a key forimproving the wet bias of summer simulation of the model.

[42] Acknowledgments. The Indian Institute of TropicalMeteorology(Pune, India) is fully funded by the Ministry of Earth Sciences, Governmentof India, New Delhi. Eric D. Maloney was supported by the Climate andLarge‐Scale Dynamics Program of the National Science Foundation undergrants ATM‐0832868 and AGS‐1025584 and by the Science and Technol-ogy Center for MultiScale Modeling of Atmospheric Processes, managed byColorado State University under cooperative agreement ATM‐0425247.Eric D. Maloney and James J. Benedict were also supported by award

Figure 16. Composite 15–30 (red line) and 35–80 (blue lines) day filtered vertical heating tendency(10−4 K s−1) of the CRM, in MMF, for active monsoon conditions averaged over the boxes shown inFigure 5b.


15 of 17

NA08OAR4320893 from the National Oceanic and Atmospheric Adminis-tration, U.S. Department of Commerce. The statements, findings, conclu-sions, and recommendations do not necessarily reflect the views of theNational Science Foundation (NSF), NOAA, or the Department of Com-merce. We thank the National Center for Environmental Prediction (NCEP)for the reanalysis data used in this paper. We thank David A. Randall(Department of Atmospheric Science, Colorado State University (CSU))for permitting us to use the model output. We also thank Mark Branson(Department of Atmospheric Science, CSU) and Daehyun Kim (Lamont‐Doherty Earth Observatory of Columbia University) for arranging the dataaccess. The MJO CLIVAR working group (http://climate.snu.ac.kr/mjo_diagnostics/index.htm) is acknowledged for the diagnostics used in someof the figures. D.W.’s contribution to this study was carried out on behalfof the Jet Propulsion Laboratory, California Institute of Technology, undera contract with the National Aeronautics and Space Administration.

ReferencesAjaya Mohan, R. S., and B. N. Goswami (2003), Potential predictabilityof the Asian summer monsoon on monthly and seasonal time scales,Meteorol. Atmos. Phys., 84, 83–100, doi:10.1007/s00703-002-0576-4.

Benedict, J. J., and D. A. Randall (2009), Structure of the Madden‐Julian Oscillation in the superparameterized CAM, J. Atmos. Sci., 66,3277–3296, doi:10.1175/2009JAS3030.1.

Chatterjee, P., and B. N. Goswami (2004), Structure, genesis and scaleselection of the tropical quasi‐biweekly mode, Q. J. R. Meteorol. Soc.,130, 1171–1194, doi:10.1256/qj.03.133.

Collins, W. D., et al. (2006), The formulation and atmospheric simulationof the Community Atmosphere Model Version 3 (CAM3), J. Clim., 19,2144–2161, doi:10.1175/JCLI3760.1.

DeMott, C. A., C. Stan, D. A. Randall, J. L. Kinter III, andM. Khairoutdinov(2011), The Asian monsoon in the super‐parameterized CCSM and itsrelationship to tropical wave activity, J. Clim., 24, 5134–5156,doi:10.1175/2011JCLI4202.1.

Duchon, C. E. (1979), Lanczos filtering in one and two dimensions,J. Appl. Meteorol., 18, 1016–1022, doi:10.1175/1520-0450(1979)018<1016:LFIOAT>2.0.CO;2.

Goswami, B. N. (1998), Interannual variation of Indian summer monsoonin a GCM: External conditions versus internal feedbacks, J. Clim., 11,501–522, doi:10.1175/1520-0442(1998)011<0501:IVOISM>2.0.CO;2.

Goswami, B. N. (2005), The Asian monsoon: Interdecadal variability, inThe Asian Monsoon, chap. 7, edited by B. Wang, pp. 295–327, Springer,Berlin.

Goswami, B. N., and J. Shukla (1984), Quasi‐periodic oscillations in asymmetric general circulation model, J. Atmos. Sci., 41, 20–37,doi:10.1175/1520-0469(1984)041<0020:QPOIAS>2.0.CO;2.

Goswami, B. N., and P. K. Xavier (2005), Dynamics of “internal” interan-nual variability of Indian summer monsoon in a GCM, J. Geophys. Res.,110, D24104, doi:10.1029/2005JD006042.

Goswami, B. N., G. Wu, and T. Yasunari (2006), Annual cycle, intrasea-sonal oscillations and roadblock to seasonal predictability of the Asiansummer monsoon, J. Clim., 19, 5078–5099, doi:10.1175/JCLI3901.1.

Goswami, B. N., M. C. Wheeler, J. C. Gottschalck, and D. E. Waliser(2011), Intraseasonal variability and forecasting: A review of recentresearch, in The Global Monsoon System: Research and Forecast, Ser.Asia‐Pac. Weather and Clim., vol. 5, 2nd ed., edited by C.‐P. Changet al., pp. 389–408, World Sci., Hoboken, N. J.

Hendon, H. H., and M. C. Wheeler (2008), Some space‐time spectral anal-yses of tropical convection and planetary‐scale waves, J. Atmos. Sci., 65,2936–2948, doi:10.1175/2008JAS2675.1.

Huffman, G. J., R. F. Adler, M. Morrissey, D. T. Bolvin, S. Curtis,R. Joyce, B. McGavock, and J. Susskind (2001), Global precipitation atone‐degree daily resolution from multi‐satellite observations, J. Hydro-meteorol., 2, 36–50, doi:10.1175/1525-7541(2001)002<0036:GPAODD>2.0.CO;2.

Hurrell, J. W., J. J. Hack, D. Shea, J. M. Caron, and J. Rosinski (2008), Anew sea surface temperature and sea ice boundary dataset for the Commu-nity Atmosphere Model, J. Clim., 21, 5145–5153, doi:10.1175/2008JCLI2292.1.

Jiang, X., T. Li, and B. Wang (2004), Structures and mechanisms of thenorthward propagation boreal summer intraseasonal oscillation, J. Clim.,17, 1022–1039, doi:10.1175/1520-0442(2004)017<1022:SAMOTN>2.0.CO;2.

Kalnay, E., et al. (1996), The NCEP/NCAR 40‐year reanalysis project,Bull. Am. Meteorol. Soc., 77, 437–471, doi:10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.

Khairoutdinov, M. F., and D. A. Randall (2001), A cloud resolving modelas a cloud parameterization in the NCAR Community Climate SystemModel: Preliminary results, Geophys. Res. Lett., 28, 3617–3620,doi:10.1029/2001GL013552.

Khairoutdinov, M., and D. A. Randall (2003), Cloud‐resolving modeling ofARM Summer 1997 IOP: Model formulation, results, uncertainties andsensitivities, J. Atmos. Sci., 60, 607–625, doi:10.1175/1520-0469(2003)060<0607:CRMOTA>2.0.CO;2.

Khairoutdinov, M., D. A. Randall, and C. DeMott (2005), Simulationsof the atmospheric general circulation using a cloud‐resolving modelas a superparameterization of physical processes, J. Atmos. Sci., 62,2136–2154, doi:10.1175/JAS3453.1.

Khairoutdinov, M., C. DeMott, and D. Randall (2008), Evaluation of thesimulated interannual and subseasonal variability in an AMIP‐stylesimulation using the CSU multiscale modeling framework, J. Clim.,21, 413–431, doi:10.1175/2007JCLI1630.1.

Kim, D., et al. (2009), Application of MJO simulation diagnostics to climatemodels, J. Clim., 22, 6413–6436, doi:10.1175/2009JCLI3063.1.

Krishna Kumar, K., K. Kamala, B. Rajagopalan, M. P. Hoerling, J. K.Eischeid, S. K. Patwardhan, G. Srinivasan, B. N. Goswami, andR. Nemanai (2011), Once and future pulse of Indian monsoonal climate,Clim. Dyn., 36, 2159–2170, doi:10.1007/s00382-010-0974-0.

Lin, J.‐L., et al. (2006), Tropical intraseasonal variability in 14 IPCC AR4climate models. Part I: Convective signals, J. Clim., 19, 2665–2690,doi:10.1175/JCLI3735.1.

Lin, J.‐L., K. M. Weickman, G. N. Kiladis, B. E. Mapes, S. D. Schubert,M. J. Suarez, J. T. Bacmeister, and M.‐I. Lee (2008), Subseasonalvariability associated with Asian summer monsoon simulated by 14 IPCCAR4 coupled GCMs, J. Clim. , 21 , 4541–4567, doi:10.1175/2008JCLI1816.1.

Liu, P., et al. (2009), An MJO simulated by the NICAM at 14‐ and 7‐kmresolutions, Mon. Weather Rev., 137, 3254–3268, doi:10.1175/2009MWR2965.1.

Luo, Z., and G. L. Stephens (2006), An enhanced convection‐wind‐evaporation feedback in a superparameterization GCM (SP‐GCM)depiction of the Asian summer monsoon, Geophys. Res. Lett., 33,L06707, doi:10.1029/2005GL025060.

Miura, H., M. Satoh, T. Nasuno, A. T. Noda, and K. Oouchi (2007),Madden‐Julian Oscillation event realistically simulated by a globalcloud‐resolving model, Science, 318, 1763–1765, doi:10.1126/science.1148443.

Oouchi, K., A. T. Noda, M. Satoh, B. Wang, S. P. Xie, H. G. Takahashi,and T. Yasunari (2009), Asian summer monsoon simulated by a globalcloud‐system‐resolving model: Diurnal to intra‐seasonal variability,Geophys. Res. Lett., 36, L11815, doi:10.1029/2009GL038271.

Pedlosky, J., and C. Frenzen (1980), Chaotic and periodic behavior of finiteamplitude baroclinic waves, J. Atmos. Sci., 37, 1177–1196, doi:10.1175/1520-0469(1980)037<1177:CAPBOF>2.0.CO;2.

Rajeevan, M., J. Bhate, J. D. Kale, and B. Lal (2006), High‐resolution dailygridded rainfall data for the Indian region: Analysis of break and activemonsoon spells, Curr. Sci., 91, 296–306.

Sato, T., H. Miura, M. Satoh, Y. N. Takayabu, and Y. Wang (2009), Diurnalcycle of precipitation in the tropics simulated in a global cloud resolvingmodel, J. Clim., 22, 4809–4826, doi:10.1175/2009JCLI2890.1.

Sikka, D., and S. Gadgil (1980), On the maximum cloud zone and theITCZ over Indian longitudes during the southwest monsoon, Mon.Weather Rev., 108, 1840–1853, doi:10.1175/1520-0493(1980)108<1840:OTMCZA>2.0.CO;2.

Slingo, J. M., P. M. Inness, and K. R. Sperber (2005), Modeling, inIntraseasonal Variability in the Atmosphere‐Ocean Climate System,chap. 11, pp. 361–397, edited by W. K.‐M. Lau and D. E. Waliser,Springer, Heidelberg, Germany.

Sperber, K. R., J. M. Slingo, D. E. Waliser, and P. M. Inness (2008),Coarse‐resolution models only partly cloudy, Science, 320, 612a–613s,doi:10.1126/science.320.5876.612a.

Stan, C., M. Khairoutdinov, C. A. DeMott, V. Krishnamurthy, D. M.Straus, D. A. Randall, J. L. Kinter III, and J. Shukla (2010), Anocean‐atmosphere climate simulation with an embedded cloud resolvingmodel, Geophys. Res. Lett., 37, L01702, doi:10.1029/2009GL040822.

Tao, W.‐K., et al. (2009), Multi‐scale modeling system: Development,applications and critical issues, Bull. Am. Meteorol. Soc., 90, 515–534,doi:10.1175/2008BAMS2542.1.

Waliser, D. E. (2006), Intraseasonal variability, in The Asian Monsoon,edited by B. Wang, pp. 203–257, Springer, New York.

Waliser, D. E., et al. (2003), AGCM simulations of intraseasonal variabilityassociated with the Asian summer monsoon, Clim. Dyn., 21, 423–446,doi:10.1007/s00382-003-0337-1.

Waliser, D., et al. (2009), MJO simulation diagnostics, J. Clim., 22,3006–3030, doi:10.1175/2008JCLI2731.1.

Wang, B. (2005), Theory, in Tropical Intraseasonal Oscillation in theAtmosphere And Ocean, edited by W. K.‐M. Lau and D. E. Waliser,chap. 10, pp. 307–351, Springer, Heidelberg, Germany.


16 of 17

Wang, B., and X. Xu (1997), Northern Hemisphere summer monsoon singu-larities and climatological intraseasonal oscillation, J. Clim., 10, 1071–1085,doi:10.1175/1520-0442(1997)010<1071:NHSMSA>2.0.CO;2.

Webster, P. J., V. O. Magana, T. N. Palmer, J. Shukla, R. A. Tomas,M. Yanai, and T. Yasunari (1998), Monsoons: Processes, predictability,and the prospects for prediction, J. Geophys. Res., 103, 14,451–14,510,doi:10.1029/97JC02719.

Wheeler, M., and G. N. Kiladis (1999), Convectively coupled equatorialwaves: Analysis of clouds and temperature in the wave number‐frequencydomain, J. Atmos. Sci., 56, 374–399, doi:10.1175/1520-0469(1999)056<0374:CCEWAO>2.0.CO;2.

Wilks, D. S. (1995), Statistical Methods in the Atmospheric Sciences,Academic, San Diego, Calif.

Yasunari, T. (1979), Cloudiness fluctuations associated with the NorthernHemisphere summer monsoon, J. Meteorol. Soc. Jpn., 57, 227–242.

J. J. Benedict and E. D. Maloney, Department of Atmospheric Science,Colorado State University, Fort Collins, CO 80523, USA.B. B. Goswami, B. N. Goswami, N. J. Mani, and P. Mukhopadhyay,

Indian Institute of Tropical Meteorology, Dr. Homi Bhabha Road,Pashan, Pune‐411008, India. ([email protected])M. Khairoutdinov, School of Marine and Atmospheric Sciences, Stony

Brook University, New York, NY 11794, USA.D. E. Waliser, Jet Propulsion Laboratory, California Institute of

Technology, Pasadena, CA 91109, USA.


17 of 17

monsoon intraseasonal oscillations as simulated by the

Documents